Neural correlates of visual attention during risky decision evidence integration

John R. Purcell, Andrew Jahn, Justin M. Fine, Joshua W. BrownNeuroImage 2021 · DOI

TL;DR

This study investigates how our brains integrate visual information about potential rewards and losses when making risky decisions, using fMRI and eye-tracking. They found that the anterior cingulate cortex (ACC) is more active when focusing on potential gains, while the posterior insula (PI) is more involved in processing potential losses, suggesting distinct roles in evaluating risky options.

What Problem Does This Paper Address?

Making decisions, especially those involving risk, is a complex cognitive process. We constantly weigh potential gains against potential losses, and this involves integrating various pieces of information over time. While previous research has identified brain regions involved in evaluating risk and making choices, it's less clear how the dynamic process of evidence integration – the gradual accumulation of information that leads to a decision – relates to visual attention. Specifically, how does where we look influence our brain's processing of risky options, and which brain areas are involved in this interplay? This paper aims to bridge this gap by examining the neural correlates of visual attention during the evidence integration phase of risky decision-making.

Methods

The researchers employed a combination of functional Magnetic Resonance Imaging (fMRI) and eye-tracking to capture both brain activity and visual attention patterns simultaneously. Participants were presented with choices between a "sure-thing" option (a guaranteed reward) and a "gamble" option (which had a potential reward and a potential penalty).

  • fMRI: This technique measures brain activity by detecting changes in blood flow. It allowed the researchers to see which brain regions were active during different parts of the task.
  • Eye-tracking: This technology precisely tracks where participants are looking on the screen. By analyzing gaze patterns, the researchers could determine how much time participants spent focusing on the sure-thing option, the potential reward of the gamble, and the potential penalty of the gamble.
  • Task Design: The experiment was carefully designed to isolate the evidence integration process. Participants had a fixed amount of time to view the options before a cue indicated they could make their choice. This delay was intended to separate the cognitive deliberation from the motor response. The values of the sure-thing and gamble options were also titrated throughout the experiment to encourage participants to choose between the gamble and sure-thing options with roughly equal frequency.

Datasets

The study used data collected from 18 healthy, right-handed participants. The raw and processed data, including functional and anatomical scans, eye-tracking data, and behavioral data, are publicly available via figshare.com. No new benchmarks were introduced; the focus was on analyzing existing data with novel experimental and analytical approaches.

Key Insights and Ideas

  • Divergent Roles of ACC and Insula: The study strongly suggests that the dorsal ACC (dACC) and the posterior insula (PI) play distinct roles in processing risky decisions. The dACC appears to be more engaged when participants focus on potential gains, while the PI is more involved when they focus on potential losses.
  • Attention Guides Neural Activity: Visual attention is not just a passive observer but actively influences neural processing. Where participants looked directly correlated with activation in key brain regions.
  • Evidence Integration is Dynamic: The brain doesn't just make a snap judgment. It seems to dynamically integrate information about potential outcomes, with different brain areas contributing to evaluating the "good" and the "bad" aspects of a risky choice.
  • Loss Aversion Reflected in PI: The posterior insula's increased activation when participants foveated on the potential loss of a gamble aligns with its known role in processing aversive outcomes and risk avoidance.

Key Results

  • Behavioral: Eye-tracking revealed that when participants chose a gamble, they spent more time looking at the potential reward (G-Reward) and penalty (G-Penalty) compared to the sure-thing (ST). Conversely, when they chose the sure-thing, they spent more time looking at the ST option.
  • Neural Activation:
    • GambleReward vs. GamblePenalty: Greater activation was observed in the dACC and premotor cortex when participants foveated on the G-Reward compared to the G-Penalty.
  • GamblePenalty vs. GambleReward: Greater activation was found in the posterior insula (both left and right) when participants foveated on the G-Penalty compared to the G-Reward.
  • Correlational Findings: Activation in a specific region of the dACC was positively correlated with the propensity to choose the gamble option, suggesting its role in integrating information that favors approach behavior despite risk.

Limitations and Future Work

The authors acknowledge several limitations:

  • Fixed Looking Time: The imposed looking time before choice might not perfectly reflect real-world decision-making, which can be more spontaneous.
  • Variability in Gamble Choices: Despite efforts to control probabilities, participants varied in the number of risky gambles they chose, which could influence neural activation. Future studies could use more rigorous methods to ensure within-subject parity between risky and safe choices.
  • Lack of Outcome Feedback: The absence of outcome feedback and monetary incentives might have made the gamble values less impactful. Future research could benefit from incentivizing task performance.
  • OFC/VMPFC Activation: The lack of significant activation differences in the orbitofrontal cortex (OFC) and ventromedial prefrontal cortex (VMPFC) was surprising, given previous findings. This might be due to signal dropout in these regions or differences in task design.
  • Computational Modeling: The authors suggest that future studies could employ computational modeling approaches (e.g., drift diffusion models) to better evaluate the role of dynamic visual attention in choice behavior and evidence accumulation.

The study opens avenues for further research into how different brain regions contribute to specific aspects of risky decision-making, particularly the interplay between attention, risk evaluation, and choice.

Summary updated

BibTeX
@article{purcell2021neural,
  author = {John R. Purcell and Andrew Jahn and Justin M. Fine and Joshua W. Brown},
  title = {{Neural correlates of visual attention during risky decision evidence integration}},
  journal = {NeuroImage},
  year = {2021},
  doi = {10.1016/j.neuroimage.2021.117979},
}
Figure